MR image reconstruction using deep learning: evaluation of network structure and loss functions

Abstract

Background: To review and evaluate approaches to convolutional neural network (CNN) reconstruction for accelerated cardiac MR imaging in the real clinical context. Methods: Two CNN architectures, Unet and residual network (Resnet) were evaluated using quantitative and qualitative assessment by radiologist. Four different loss functions were also considered: pixel-wise (L1 and L2), patch-wise structural dissimilarity (Dssim) and feature-wise (perceptual loss). The networks were evaluated using retrospectively and prospectively under-sampled cardiac MR data.Results: Based on our assessments, we find that Resnet and Unet achieve similar image quality but that former requires only 100,000 parameters compared to 1.3 million parameters for the latter. The perceptual loss function performed significantly better than L1, L2 or Dssim loss functions as determined by the radiologist scores.Conclusions: CNN image reconstruction using Resnet yields comparable image quality to Unet with 10X the number of parameters. This has implications for training with significantly lower data requirements.Network training using the perceptual loss function was found to better agree with radiologist scoring compared to L1, L2 or Dssim loss functions.

Fast and accurate calculation of myocardial T1 and T2 values using deep learning Bloch equation simulations (DeepBLESS)

“…Although DeepBLESS achieved almost instantaneous T 1 /T 2 map reconstruction for the radial T 1 ‐T 2 mapping sequence, the compressed‐sensing reconstruction took approximately 3 minutes, a limitation for using the radial T 1 ‐T 2 sequence for simultaneous myocardial T 1 and T 2 mapping. Recent studies have shown that deep learning can be applied to replace compressed sensing, to reduce reconstruction time 35‐37 . These techniques may be combined with our proposed T 1 calculation technique to further reduce total imaging time and enable online use of the radial T 1 ‐T 2 mapping technique.…”

Retrospective respiratory motion correction in cardiac cine MRI reconstruction using adversarial autoencoder and unsupervised learning

“…B (from top to bottom), sample of the original motion-free image, the synthesized respiratory motion-corrupted image, and the error map between them mentioned groups, which pairs had statistically significant difference? To answer these questions, Friedman's two-way analysis 51,52 and nonparametric paired comparison tests were applied. Significance level for all statistical tests was assumed at α = 0.05.…”

A functional artificial neural network for noninvasive presurgical evaluation of glioblastoma multiforme prognosis and radiosensitivity profiling

“…The parameters that we chose to build the neural networks should stay consistent throughout our trials. A loss function is required in many neural networks and can lead to performance increases/decreases depending on the type of problem that is being tackled [24]. Figure 5A).…”